Stochastic In-Face Frank-Wolfe Methods for Non-Convex Optimization and Sparse Neural Network Training

Abstract

The Frank-Wolfe method and its extensions are well-suited for delivering solutions with desirable structural properties, such as sparsity or low-rank structure. We introduce a new variant of the Frank-Wolfe method that combines Frank-Wolfe steps and steepest descent steps, as well as a novel modification of the "Frank-Wolfe gap" to measure convergence in the non-convex case. We further extend this method to incorporate in-face directions for preserving structured solutions as well as block coordinate steps, and we demonstrate computational guarantees in terms of the modified Frank-Wolfe gap for all of these variants. We are particularly motivated by the application of this methodology to the training of neural networks with sparse properties, and we apply our block coordinate method to the problem of ${\ell }_1$ regularized neural network training. We present the results of several numerical experiments on both artificial and real datasets demonstrating significant improvements of our method in training sparse neural networks.

References in corpus (15)

• Pruning Convolutional Neural Networks for Resource Efficient Inference
• An Analysis of Deep Neural Network Models for Practical Applications
• On the Global Linear Convergence of Frank-Wolfe Optimization Variants
• Block-Coordinate Frank-Wolfe Optimization for Structural SVMs
• Learning Sparse Neural Networks through $L_0$ Regularization
• Convergence Rate of Frank-Wolfe for Non-Convex Objectives
• Coordinate Descent Converges Faster with the Gauss-Southwell Rule Than Random Selection
• Forward - Backward Greedy Algorithms for Atomic Norm Regularization
• Convergence guarantees for a class of non-convex and non-smooth optimization problems
• Sparse Activity and Sparse Connectivity in Supervised Learning
• Structured Nonconvex and Nonsmooth Optimization: Algorithms and Iteration Complexity Analysis
• Blended Conditional Gradients: the unconditioning of conditional gradients
• Convergence to Second-Order Stationarity for Constrained Non-Convex Optimization
• Conditional gradient type methods for composite nonlinear and stochastic optimization
• Generalized Stochastic Frank-Wolfe Algorithm with Stochastic "Substitute" Gradient for Structured Convex Optimization